geo-mcp
OSM-backed multimodal navigation MCP server enabling geocoding, route planning, POI discovery, and urban layout analysis. It provides a structured city graph to LLM apps without proprietary map APIs.
README
geo-mcp
OSM-backed multimodal (walk + transit) navigation MCP server. Gives LLM apps a structured graph of the city so they can geocode, route, explore POIs, and reason about urban layout without proprietary map APIs.
Features
- Clean Architecture / DDD (Python 3.14): domain pathfinding stays free of OSM/NetworkX details
- Multimodal graph: OSMnx walking network + OSM transit stops/routes + transfer edges
- Disk cache: bbox-keyed local graph cache under
~/.cache/geo-mcp/ - MCP tools compatible with Cursor, Claude Desktop, Codex, etc. (stdio)
Tools
| Tool | Purpose |
|---|---|
load_area |
Fetch/cache walk+transit graph for a place or bbox |
geocode |
Place / address → coordinates |
reverse_geocode |
Coordinates → place |
plan_route |
Multimodal A* path with legs + narrative |
find_nearby |
Nearby OSM POIs |
list_transit_options |
Nearby stops and lines |
describe_area |
Major roads, neighborhoods, landmarks, connectivity notes |
snap_to_network |
Nearest graph node to a coordinate |
Requirements
- Python 3.14+
- uv
Install
cd /path/to/geo-mcp
uv sync
Run (stdio MCP)
uv run geo-mcp
Cursor config
Project example: .cursor/mcp.json
Or add to ~/.cursor/mcp.json:
{
"mcpServers": {
"geo-mcp": {
"command": "uv",
"args": [
"--directory",
"/path/to/geo-mcp",
"run",
"geo-mcp"
]
}
}
}
Suggested agent flow
geocodethe destination (and origin if needed)load_areafor the neighborhood / placedescribe_areafor spatial contextplan_routewith coordinates- Optionally
find_nearby/list_transit_options
Tests
uv run pytest
Live OSM smoke (needs network):
uv run python scripts/smoke_route.py
Configuration
Environment variables use the GEO_MCP_ prefix:
GEO_MCP_USER_AGENT— Nominatim/Overpass User-AgentGEO_MCP_CACHE_DIR— graph cache directoryGEO_MCP_NOMINATIM_URL/GEO_MCP_OVERPASS_URL
Attribution & policy
- Map data © OpenStreetMap contributors
- Respect Nominatim usage policy and Overpass etiquette (polite User-Agent, modest request rates)
Architecture
src/geo_mcp/
domain/ # entities + pure pathfinding
application/ # use cases + DTOs
infrastructure/ # OSM adapters + cache
interfaces/mcp/ # FastMCP tools
MVP limits
- Transit uses OSM topology / heuristic speeds (not GTFS schedules)
- Large city-wide bboxes are automatically shrunk around the geocoded center
- First
load_areafor a bbox can take a while while Overpass/OSMnx respond
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。